owwzo/ai-agent/gpt-image-2-style-library

GPT-Image2 Style Library

Use this skill to turn a user's image-generation intent into a production-ready GPT-Image2 prompt using the awesome-gpt-image-2 style library.

Example Output

[Image: City life system map example]

Example request: 用 gpt-image-2-style-library 技能生成城市生命系统图谱

Reference

• Read references/style-library.md before choosing a template or style.
• The reference is generated from data/style-library.json in the repository.
• Prefer the reference over memory when template names, categories, covers, or style tags matter.

Workflow

1. Detect the user's language and answer in that language.
2. Identify the user's target output: product, poster, UI, infographic, brand, photo, illustration, character, scene, history, document, or special task.
3. Match the request in this order: template category, visual style tag, scene tag, then nearest example cases.
4. If one template is clearly strongest, use it directly. If several are plausible, present 2-3 options with short reasons and ask the user to choose.
5. Build the final prompt with these blocks:

• subject and task
• composition and layout
• visual style and materials
• text and label requirements
• aspect ratio and output format
• constraints and negative details

1. Include the selected template name and any useful example case IDs.

Reference-image character poster workflow

For a user-uploaded character image intended for a poster:

1. Inspect the image first and extract identity anchors: face shape, hairstyle, eye color/shape, costume silhouette, palette, props, pose, line/rendering style, and emotional tone.
2. Treat the uploaded image as the primary source for appearance consistency. Use web/reference research only to supplement character background, symbolism, mood, and exact copy; do not override visible traits with conflicting fan-wiki descriptions.
3. If the user asks for a plan first, do not generate an image. Return the extracted anchors, proposed template, composition, palette, text hierarchy, aspect ratio, and consistency/negative constraints for approval.
4. For known fictional characters, verify the meaning and provenance of requested taglines. Distinguish chapter titles, fan slogans, and canonical quotations instead of presenting uncertain wording as canon.
5. After approval, build the prompt with explicit identity anchors and limit poster copy to the exact requested title/subtitle to reduce text artifacts.

Output Defaults

• Provide a copyable prompt first.
• Keep constraints concrete: exact text, aspect ratio, readable labels, layout hierarchy, and avoided artifacts.
• For Chinese requests, write the final prompt in Chinese unless the user asks for English.
• For English requests, write the final prompt in English unless the user asks for Chinese.
• When the user asks for multiple concepts, reuse one template and vary subject, composition, palette, and scene.

Maintenance

When the source repository changes, run:

npm run generate:style-skill

To install the skill into the local Codex skill folder, run:

npm run install:skill